LOADS – AN IMPORTANT TASK FROM PRE-DESIGN TO LOADS FLIGHT TESTING

T. Klimmek1, P. Ohme2, P. D. Ciampa3, and V. Handojo1, DLR, German Aerospace Center, 1Institute of Aeroelasticity, 37073 Göttingen, Germany, 2Institute for Flight Systems, 38108 Braunschweig, Germany, 3Institute of Air Transportation Systems, 21079 Hamburg

Summary The estimation of loads acting on an aircraft structure is an indispensable task ranging from conceptual, preliminary, and detail design to loads flight testing when an aircraft is already in service. Work package 4 of the DLR project iLOADS covers the range broadly. Physics-based loads and aerodynamic performance analysis methods in pre-design are presented using a parameterized and automatized process nested in an overall aircraft design loop. Such process can be used for conventional an unconventional configurations. Covering the complete design process, a parametric set-up for structural modeling, estimation of loads and loads flight testing for the DLR HALO, a modified Gulfstream G550, is also shown. Finally the in-flight system identification activities including flight tests for the DLR Discus-2c are described.

or analytical methods are applied, because knowledge of 1. INTRODUCTION the design of a new aircraft is rare or wide range parameter studies require fast analyses methods [15]. To Aircraft loads analysis is an important engineering improve the overall aircraft design (OAD) process physics discipline playing a major role from the first sketch of a based methods for aeroelastic loads analysis and the new aircraft until the decommissioning. A proper aerodynamic performance assessment of the flexible estimation of the loads in the design process is reflected structure are integrated in an OAD process shown in by the structure withstanding the loads and essentially by chapter 2. the structural weight that is about to be minimized. The monitoring of the loads acting on the structure when the The loads analysis for the complete aircraft is shown for aircraft is in service is as well significant. On the one hand the DLR HALO, a slightly modified G550 from Gulfstream the measured loads are compared to the predicted loads Aerospace Company (GAC) to be used for atmospheric and on the other hand the loads data can be used to research. The DLR HALO lacks an applicable structural estimate the real lifetime of an aircraft mainly driven by model for loads analysis. Such structural models are fatigue. normally company confidential and consequently publicly not available. Therefore, one task of iLOADS was to set A big portion of such a wide range is covered by the up a structural model for the DLR HALO to enable loads research activities at DLR within work package 4 of the analysis. The process, wherein limited available data from DLR project iLOADS (2013-2016). The project iLOADS is GAC were used, and the results of selected load cases is focused on the loads process and has the task to gather presented in chapter 3. almost all research activities respectively research institute, being part of or closely connected to the aircraft Regarding the measurement of flight loads two DLR loads process. A more detailed description of the project aircraft were considered, on the one hand again the DLR is given in [10]. In work package 4 of iLOADS a wide HALO, with the focus on the interface loads between the range of loads analysis and measurement is covered from stores and the , and on the other hand the DLR conceptual design, and loads analysis for a complete Discus 2c, where the measurement of loads is part of an aircraft to loads flight testing. overall system identification. Few publications are available where a comprehensive Like the DLR HALO is also designed to survey regarding loads analysis is presented. The books carry wing stores, for the DLR HALO to be used for from Lomax [11], as well as the one from Wright and atmospheric research. As the manufacture sets limiting Cooper [16], are describing various loads analysis loads between the stores and the aircraft, it has to be methods for almost all load types that have to be shown in advance for new stores or already certified considered in aircraft design. Niu presents in his book [13] stores with another mass and center of gravity that the on stress analysis besides basic loads analysis interface loads are not exceeded. To improve the methods, the interfaces and the interdependencies of the capabilities of DLR to support such certification process, loads group to the groups and departments of a typical the flight tests were done (see chapter 4). aircraft manufacture. Finally, as part of a comprehensive system identification The estimation of loads acting on an aircraft is an flight test of the sailplane DLR Discus-2c were executed. important task already at early steps of the design, The measurement equipment is shown, as well first because the structural mass, always to be minimized, is results, that are planned to be used to set up a loads the result of a structural sizing with the assumed design model purely based on flight-test (see chapter 5). loads. In conceptual design normally only statistics based

2. LOADS ANALYSIS IN PRE-DESIGN

2.1. Loads in Overall Aircraft Design

The early aircraft development stages focus on the feasibility and on the assessment of the overall performance of the investigated configurations. During the pre-design stages design details are typically not available, and the overall aircraft synthesis relies on the specification of TLAR (Top Level Aircraft Requirements), such as transportation mission and operational constraints, and provides as output the overall aircraft parameters, such as MTOW (Maximum Take Off Weight), mission fuel consumption, etc.[21].

Despite the large impact of loads on the aircraft performance, conceptual overall aircraft design (OAD) methodologies are mainly based on statistical data. At best, the airframe’s loads carrying components are estimated by empirical relations or database available from previous designs rather than by incorporating physics-based analyses. Furthermore, at the early stages the estimation of the relevant load cases includes conservative allowances and safety margins for the structural design in order to guarantee the structural integrity. This rather conventional design approach usually adds an “aeroelastic penalty” [4] to the final designed structures, which may be unveiled only at the later phases of the development. Flexibility effects on the performance are also conventionally not included during the pre-design phases of the investigations. However, with the increasing efficiency of the structural design concepts, the importance of flexible effects is significantly growing. Those effects need to be properly accounted from the beginning of the aircraft development in order to minimize FIGURE 1. Multi-Fidelity Loads Process expensive redesign activities or the degradation of the prospected performance. The recent advancement in The first phase is the initialization of the aircraft, labeled in computational performance and simulation capabilities the schema as “L0 conceptual”. The initialization step provide accessibility to time efficient physics based provides the first full aircraft synthesis for a given set of models which could be adopted from the beginning of the TLAR. The synthesis is provided by a conceptual aircraft developments. synthesis module, whose formulations are based on typical aircraft design methodologies. The output is a The integration of such models in overall aircraft synthesis representation of synthesized aircraft, and its process is here presented, and it is part of an overall performance, such as the aircraft design masses, mission framework aiming to enhance the aircraft pre-design fuel as shown in FIGURE 2. Further, the solution is activities of conventional and unconventional augmented by additional features in order to enable to configurations. Focus in on accounting a low fidelity loads progress the design solution to the next phase which is analysis process and investigate the impact it has on the using low fidelity physics based analysis modules. early design stages. The full automation of such a process Examples of enhanced details are the initialization of a is a key feature, since at the pre-design stages large representative wing internal structural topology, the three trade-off studies, and design of experiments are expected dimensional shape of the aircraft components. to be performed.

2.2. Multi-fidelity Loads Process

A fully automated OAD multi-fidelity process is assembled within the iLOADS Project. The process is characterized by 3 phases, with focus on the aero-structural design modules which are used to investigate how a physics loads analysis process may impact the overall aircraft synthesis and the designed performance. A representation schema of the process is shown in FIGURE 1. FIGURE 2. Initialization Phase – Conceptual L0

Thereafter, a mission analysis is performed, and an Thereafter, the solution is forward to the second phase, in updated value of the mission fuel is calculated. Respect to which the physics based low fidelity modules analyze the the mission fuel calculated in the first phase, the updated initialized aircraft. The second phase is labeled in the amount account for the updates structural masses and schema as “L0 conceptual + L1 physics based”. Although flexible performance which are provided by the physics the approach may be extended to comprise all the aircraft based simulations. disciplines, in the current process the focus is on the aero-structural analysis and design modules. The aim is to provide a more accurate representation of the physics behaviors with respect to the previous conceptual phase. At this stage two main effects are identified and investigated by making use of the modules:

1. The aero-structural sizing loop: in this sub- process the primary structural components are sized with respect to a selected set of critical loads cases. The output consists in the sized structures and the corresponding structural masses. FIGURE 3. Multi-fidelity, physics based disciplinary modules 2. The flexibility loop: the aircraft performance (e.g. The third stage is here labeled as “multi-fidelity synthesis”. the aerodynamics polar) are evaluated considering All the updated aircraft characteristics are used to update the flexibility effects of the airframe, and a thus the OAD synthesis and the overall estimation of the evaluating the static aeroelastic equilibrium for each performance. In this phase the conceptual design tool is flow condition. The FSI coupling is taken into account still the one called by the initialization phase, but values to determine the lift and drag coefficients of the calculated from the physics based modules replace the aircraft, for relevant combinations of angle of attack, conceptual estimations. The updated synthesis will Mach and Reynolds number. Hence, the updated produce new overall aircraft design values (MTOM), an aircraft aerodynamic performance, corrected by the updated initialized aircraft become available. The solution flexibility effects can be progressed once more to the physics based phase. Hence, the full multi-fidelity cycle is iteratively For both the effects, an available VLM based module is repeated via the fixed point iteration architecture, until the used to estimate the aerodynamics efficiency at various convergence of the aircraft design masses. conditions of the flight envelope, and to provide the aerodynamics loading distribution on the lifting surfaces The integration of the described process is a workflow as resulting from the critical design maneuvers. The assembled in RCE environment [6] [18], whose structural representation relies on a simplified Finite corresponding workflow is shown in FIGURE 4, and the Element (FE) formulation for the modeling, analysis and transfer of data between modules is via the CPACS post-processing of the aircraft primary structures (i.e. format [5]. All the models generations, analysis and explicit wing-box modeling), in order to determine the design are fully automated. Further, the process may be displacements and the stress fields of the aircraft under employed to analyze any type of configuration described multiple load cases. A number of sizing strategies, such in the CPACS file (within the limits of the physics fully stress design, and flexural buckling criteria, are formulation behind the modules). Examples of models implemented for the dimensioning of the selected primary generated by the system for unconventional aircraft are structures. In order to account for the aircraft flexibility shown in FIGURE 5. effects, the fluid-structure interactions (FSI) need to be considered in the aero-structural analysis and sizing process. The aero-structural coupling is implemented by first mapping the aerodynamics forces on to the structural model, and then transferring the computed displacements on the structural nodes to the aerodynamic geometry. Since the disciplinary modules are loosely coupled, the coupling schemes implemented are based on MLS\RBF for the forces-displacements transfers in both the sizing and the flexibility loops [2].

FIGURE 3 shows the disciplinary models generated by the analysis tools, namely the aerodynamics VLM lattice for the lifting surfaces, and the FE beam model of the aircraft. The nodal deflections are also shown for the main wing, under the critical sizing load case. Figure 3 shows the structural model with the aerodynamics loads on the FE nodes of the main wing, as resulting from the mapping schema from the VLM lattice to the structural grid, and the structural nodal displacements of the FE model and the propagation of the displacements on the geometry, via FIGURE 4. Multi-Fidelity Loads Process RCE Workflow mesh deformation techniques available in the module, applied directly on the initial geometry, and on the disciplinary VLM grid Comparing the results of the L0 conceptual synthesis with the reference, and the VAMP project, they are all very close. It has to be noted that both the conceptual system and the VAMP system, were calibrated against the reference values.

The results from the multi-fidelity approach refer to a fully converged synthesis (typically 3 or 4 fixed point iterations). The number of critical loads cases included for the structural sizing range from 1 (a representative 2.5 g FIGURE 5. Disciplinary modules automation for maneuver), to 6 steady maneuvers within the design unconventional CPACS configurations envelope, to 10 (which includes additional gust loads). 2.3. Use case The converged aircraft design masses show a difference between the L0 conceptual case, and the multi-fidelity A conventional transportation aircraft configuration, the which includes the physics based analysis. The main DLR D-150, is selected as use case of the multi-fidelity difference is in the operating empty mass (OEM) values, process described in the previous section. Among the resulting by an under estimation of the computed others, the main mission’s requirements are a design structural masses. For a conventional configuration, range of 3500 nm, at Mach 0.78, with 150 passengers. conceptual design tools can provide very accurate results, Although the set of TLAR is sufficient for the conceptual since extensive database are available, and the synthesis synthesis, additional tools’ specific inputs are required for process is calibrated on real aircraft data. On the other the other disciplinary modules, e.g. materials allocation, hand, physics based analysis would need to account for selection of the propulsion system technologies. The the simulation of a multitude of critical flight conditions process described in the previous section can be and phenomena, to produce accurate results, without executed with the following modalities: only conceptual calibration factors. However, TAB 2 shows than increasing design process (L0 level), conceptual and physics based the critical loads cases has a mitigating effect, and (multi-fidelity L0 + L1 level) with and without flexibility physics design drivers are captured within acceptable effects. In this, way the designer can tailor the process time. according to required level of accuracy and/or computational speed. A second study is setup by adding the flexibility loop and On the conventional aircraft use case two incremental the effects on the performance evaluation. In this case the studies are performed. The first study focus on the sizing representative number of sizing loads cases is reduced to loop, and the sensitivity it has on the OAD synthesis. The 2. TAB 3 shows the results for the multi-fidelity with and comparison is between the synthesis results from the only without the flexibility loop enabled. For the rigid case the conceptual process, against the multi-fidelity synthesis differences refer to the only conceptual L0 solution. Once presented (but without the flexibility loop at first). Further, more the under estimation of the structural component is the multi-fidelity approach is performed multiple timed with stemming from the limited amount of critical conditions a varying number of critical loads cases to be included in considered in the process. The reported flexible results the loads process for the structural sizing task. The show the differences of the synthesis with respect to the synthesis results are also compared against the reference rigid case. For an aircraft featuring a conventional swept- data available for the aircraft model, and additionally back wing system with moderate aspect ratio, the against a previous internal project VAMP [12], in which the structural flexibility is known to result into a degradation of same aircraft was design by making use of higher fidelity the aerodynamics performance respect to the rigid disciplines. TAB 1 and TAB 2 presents the comparison of analysis, as shown as well by the results in TAB 3. In fact the aircraft synthesis design masses and performance for the flexibility effect, when propagated through the OAD all the mentioned cases. loop, generates an increase in fuel mass, and OEM in order to satisfy the defined TLAR TAB 1. D150 synthesis comparison TAB 3. Multi-fidelity: Loads impact on flexible performance

TAB 2. Multi-fidelity: Loads impact on sizing Both the studies show the impact the loads process has on the OAD synthesis at the early development stages. The impact is even larger for the investigation of unconventional configurations for which the conceptual design phase may not rely on an existing database. Hence, in order to capturing the major design drivers, physics based modules need to be included to unveil the driving phenomena.

hundred load cases for representative station of the 3. SET-UP OF A STRUCTURAL MODEL FOR aircraft and the corresponding parameter for each load LOADS ANALYSIS FOR DLR-HALO case (e.g. flight point, mass configuration, trim or gust analysis parameter). For the outer geometry of the DLR 3.1. Research Aircraft DLR-HALO HALO a geometry model originally prepared for CFD calculations at DLR was used. Finally the results of the ground vibration test were taken to compare the structural The DLR HALO (High Altitude and LOng Range) dynamic characteristics of the structural model. Research Aircraft is an aircraft for atmospheric research and earth observation of the German science community. The HALO is based on a production G550 business jet 3.3. DLR-AE MONA Process from Gulfstream Aerospace Cooperation (GAC). For research purpose external wing stores at three stations For the set-up of the structural model the DLR-AE MONA per side can be supported by the DLR-HALO. The stores process was applied. MONA stands for the two main comprise pod, pylon, and hanger beam. All external computer programs that are used in the process, ModGen stores require a separate certification. In order to and MSC Nastran. MONA is a combined loads and design strengthen the certification capabilities of DLR regarding process. The loads estimated in the first step are used for the use of new wing pods, a specific DLR HALO loads the application of various dimensioning methods like e.g. process was established within the DLR project iLOADS. structural optimization with optimization algorithms to It is planned for the future to use such loads process to estimate the structural dimensions (e.g. wall thickness, support the certification process for new wing pods, like area of stiffener elements) of each aircraft component. the Particle Measurement System, the so-called PMS carrier (see FIGURE 6). ModGen is a DLR-AE in-house program to set up all simulation and optimization models for the loads analyses and the structural sizing of an aircraft component or the complete aircraft configuration. The model set-up is based on a parametric approach. This allows for the set-up of structural models where almost all structural parts are modelled according to the construction while a wide range of variations are still possible (e.g. outer geometry, construction concept). MSC Nastran is used for all simulation and the structural optimization task of the MONA process.

FIGURE 6. DLR-HALO with mounted PMS carrier

3.2. DLR HALO Data for the Structural Model and the Loads Analysis

Mandatory for a classical loads process is the use of representative structural models, comprising the stiffness characteristics of the aircraft and various mass configurations for a wide range of payload and fuel variants. As far as DLR lacks a structural model of the DLR HALO, several available reports and data from GAC FIGURE 7. DLR-AE MONA process adapted to DLR- and DLR that could contribute to the set-up of such model HALO were taken into account. The structural model was planned to be realized as MSC Nastran finite element For the set-up of the structural model for the DLR HALO a model. modified MONA process is established as shown in FIGURE 7. The first step of MONA is the parametrization of the geometry and the structural design of each Parts of the maintenance manual from GAC were used to component with ModGen. Necessary data from GAC and model the load carrying structure of the wing-like DLR (e.g. geometry, construction concept) are taken into structural components appropriately. Thus the wing and account. For the wing-like components the structural the tail structure are covered. Furthermore, GAC made a dimensions are estimated by a preliminary cross section mass model for the minimum equipped weight (MEW) sizing (PCS) method, an empirical-analytical method, configuration available and center of gravity positions for using the available cutting loads from GAC (see also [3] various fuel conditions. Rules for the adding of mass and [9]). Transformed to the load reference axis items to the mass model were given as well. Regarding the loads, GAC provided DLR with the loads coordinate system, the bending moment Mx, the torsion report used for the certification of the DLR HALO. The moment My, and the vertical force Fz are used to size the loads report contains a set of cutting loads of several thickness of the skin, the spars, and the ribs as well as the dimensions of the stringer and the caps. Inner stiffener elements of the spars and ribs were sized as well. For defined flight conditions (e.g. altitude, Mach according to [22]. The fuselage is modelled with beam number, load factor) and trim conditions (horizontal tail elements. For the sizing the method described in [1] is with fixed angle, angle is free trim variable), MSC used. Therein the diameter of selected fuselage sections Nastran SOL144 simulations were performed to estimate including the consideration of stiffener elements, like maneuver loads. The resulting bending moment Mx for the stringer, is taken into account. For the mass model on the wing in spanwise direction is shown in FIGURE 9. one hand the MEW mass model from GAC is used. On the other hand the mass models for the fuselage payload 1 and various fuel configurations are set up by DLR-AE. The GAC distribution of the payload mass items is done according 0.8 DLR x

to the rules given by GAC in a mass report. The fuel mass a m modeling is done also with ModGen, where the individual , 0.6 x

filling levels and deck angles of the aircraft can be M / considered for each . x 0.4 M 0.2 3.4. DLR-HALO Finite Element Model 0 The resulting finite element model is shown in FIGURE 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 8a. For the loads analysis the structural model is y/0.5b [ ] condensed to selected nodes on the load reference axis (LRA) of each component (see FIGURE 8d). The condensation is based on the interpolation of the 6 FIGURE 9. Normalized bending moment Mx for the wing degrees of freedom (DOFs) of the LRA grids by 6 for a 2.5 g symmetric maneuver load case selected DOFs of the fe grids of the full structural model shown in FIGURE 8a. Therefore the MSC Nastran The maximum deviation of the estimated loads compared element RBE3 with UM option is used. All distributes to the GAC loads for the specific load case is about 8% mass items (see FIGURE 8c) are also connected to the the average deviation about 3%. It is assumed that the LRA grids. For the aerodynamics, a panel model for the differences are mainly caused by the aerodynamic model. Doublet Lattice Method (DLM) of MSC Nastran is set up A correction of the DLM model with spanwise available by the MONA process where only the lifting surfaces are aerodynamic coefficients based on wind tunnel considered (see FIGURE 8b). In order to take wing twist measurement were not yet taken into account. and camber into account, corresponding correction Furthermore the aerodynamic characteristics of the factors are estimated by ModGen (MSC Nastran matrix fuselage were not considered. W2GJ). 1 GAC

] 0.8 DLR

[

x

a 0.6 m , y M

/ 0.4 y M 0.2

0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x/l [ ]

FIGURE 10. Normalized bending moment My along the fuselage for a 2.5 g symmetric maneuver load case FIGURE 8. Finite element model HALO-S and The resulting bending moment M along the fuselage is aerodynamic model for DLM y shown in FIGURE 10. The difference between the loads The engine mass is connected to the fuselage beam with estimated with the MONA process and the GAC loads for beam elements representing the pylon. The stiffness of the maximum load at the wing fuselage intersection is the pylon beams is oriented to the results of the GVT test about 4%. The average difference for all monitor stations for the DLR HALO performed in 2009. is about 2%. As preparation for the static tests of the PMS carrier mounted on the vibration table MAVIS of DLR-AE, 3.5. Results Loads Analysis a gust load analysis of a load case from GAC was performed with MSC Nastran Sol146. The simulation was For the loads analysis various maneuver and gust load done with the complete aircraft model with a wing pod cases were selected from available GAC load data. The mounted at the mid station. The time history of the structural model was set up according to the given data. displacements, velocities, and accelerations of a For the aircraft without fuel mass the mass and the center reference point near the LRA at the mid wing station was of gravity position are given. For the fuel configuration the used as input signal for the MAVIS to simulate vibrations fuel mass, the presumptive fuel distribution between the acting on the PMS carrier that are caused by a gust wing tanks and the deck angle of the aircraft are defined encounter. ]

4. LOADS FLIGHT TESTING WITH DLR HALO –

[ 1

t n

In April 2016 five test flights with DLR HALO were carried e out at the DLR site Oberpfaffenhofen, with around 13 m c e a

flying hours in total. The aims were to collect flight loads l 0.5

measurement data at the wingpods to validate DLR s p i d simulation methods for loads analysis and to investigate d the aircraft oscillations in flight using an online monitoring e m

system. The latter is described further in the paper by r o

Sinske [19]. The utilized wingpods are PMS (Particle n 0 Measurement System) carriers which are mounted on the 0 1 2 3 4 time [s] middle wing station on each side of the wing, as shown in FIGURE 11. FIGURE 13. Time history of the normed hardpoint displacement from a numerical gust simulation

4.2. Strain gage calibration and pre-flight tests

The flight loads measurement for the PMS carrier is performed with strain gages [19] [20]. The calibration of the strain gages is done by applying linearly independent loads on the PMS carrier and measuring the strain signals. With sufficient applied loads – more than the number of bridges – the relation between load and strain signal can be identified which enables the calculation of loads based on strain signals [20]. FIGURE 11. PMS carrier mounted on the wing station The first calibration of the strain gages was carried out at 4.1. Gust simulation on the vibration table DLR in Göttingen. The load application was conducted MAVIS with the help of sandbags. Weights up to 50 kg were attached to each force transmission point. To minimize error due to hysteresis, the applied loads were increased One preliminary test for the flight experiment was a gust and decreased step by step. The strain signals were simulation of the PMS carrier at DLR in Göttingen. For this measured for each step. A typical strain response of a aim the PMS carrier was mounted upside down on the two-step load case is shown in FIGURE 14. vibration table MAVIS (Mehrachsen-Vibrationssimulator, multi axis vibration simulator). The setup is illustrated in FIGURE 12. The gust load case selected for the simulation is a design load case for the PMS carrier, with a gust gradient of 50 feet (16 m). In order to generate the 1 input signal for the MAVIS, a numerical gust simulation ] − [ with models explained in section 3 was performed n i

beforehand. In the numerical simulation the PMS carrier a r was represented by rigid bodies and the hanger beam by t 0.5 s an elastic beam. The results in form of deflection, d velocities and accelerations were extracted, scaled down e m and taken as inputs for the gust simulation on the MAVIS. r 0 o

The time history of the hardpoint displacement resulting n from the numerical simulation is depicted in FIGURE 12. 0 100 200 300 time [s] Force transmission FIGURE 14. Typical strain response of a two-step load PMS-Carrier application Hanger beam Plate adaptor with The second calibration of the strain gages was performed Piezo scale at DLR in Oberpfaffenhofen where both PMS carriers Double T- were mounted on the wing station of the DLR HALO. In beam total nine linearly independent loads were applied on each PMS carrier, and transfer matrices to calculate loads from strain signals were derived. A picture of load application during the calibration is presented in FIGURE 15. FIGURE 12. Strain gage calibration setup on vibration table MAVIS In addition, an electromagnetic compatibility test and a increases. This indicates that the strain reacts in phase taxi vibration test were carried out beforehand to ensure with the roll acceleration. A further investigation of the the functionality and the safety of the measurement loads based on strain responses is still to be done. systems in flight [19]. ] − [ 0.5 1 s e

n 2 o 3 s p 0 e r

4 n i

a 5 r −0.5 s t 6 d

e 7 m r −1

o 8

n 975 980 985 990 995 1000 1005

] time [s] g

e 50 d [

e l g 0 n a k n

a −50 b 975 980 985 990 995 1000 1005

time [s] FIGURE 15. Strain gage calibration for one PMS carrier FIGURE 17. Normed strain signal response during a roll 4.3. Flight tests maneuver Beside strain responses, atmospheric turbulence data The second flight was conducted for loads measurement, was also collected during the flights. Its first analyses in where several maneuvers were flown for different the frequency domain show that the power spectral conditions. In total there are 21 flight conditions consisting density of the measured turbulence is similar to a von- of seven altitudes ranging from 12000 ft to 35000 ft Kármán spectrum [7], however with a smaller scale of (3658 m to 10668 m) and three speeds each. The turbulence than 2500 ft (762 m) which is defined in CS25 maneuvers for each flight condition consist of impulsive [14] A smaller scale of turbulence means that more inputs on the and pedals, a roll up to ± 45° energy of the turbulence is contained in oscillations with and a pull-up up to 2 g. higher frequencies. An example of the power spectral density of the measured turbulence and reference von- The initial aim was to simulate gust loads using Kármán spectra are depicted in FIGURE 18, where L maneuvers since gust loads are design loads for the stands for the scale of turbulence. With this finding, loads hardpoints where the hanger beam is mounted. However, analysis methods regarding atmospheric turbulence can to do so stick raps with a frequency of up to 14 Hz would be improved for the future. be necessary. Such procedure is difficult to perform with high precision. Furthermore, the control surface actuation system filters the high frequency input to some degree.

Instead, impulsive inputs on the steering system and low ] z H frequent maneuvers as described above were considered / −5 2

s 10 as safer and easier to perform. A time history of the load / 2

factor during a pull-up maneuver is presented in FIGURE m [ 16. D S P

−10 2 10 −2 0 ] 10 10 − [

1.5 frequency [Hz] r o t

c measured

a 1 f von Karman, L=76.2 m d a von Karman, L=254 m o 0.5 l von Karman, L=762 m (CS25) 0 0 5 10 15 20 FIGURE 18. Power spectral density of measured time [s] turbulence in comparison with reference von- Kármán spectra FIGURE 16. Load factor during a pull-up maneuver A normed strain signal response of the left hand side PMS carrier with all eight channels during a roll maneuver is depicted in FIGURE 17. It is evident that the strain response is already detectable before the bank angle 5. SYSTEM IDENTIFACTION FOR DLR DISCUS 2C

This iLOADS work package focuses on the development of real time capable rigid-body and flexible multipoint loads models based on flight test data using System- Identification in time domain [8]. As test aircraft the new DLR Discus-2c sailplane was used (see FIGURE 19). In addition to a basic flight test instrumentation including IMU, nose boom with 5-hole probe and control surface deflection sensors several strain sensors and accelerometers within wing, fuselage and horizontal tail were installed already during the aircraft construction. The strain sensors were calibrated in an extended ground test campaign to enable local loads measurements [23]. FIGURE 20. Structural and aerodynamic loads calculated These direct loads measurements at different positions of from 46 strain gauge measurements [1] the aircraft structure were used to extend the typical set of observation variables for enabling the prediction of local aerodynamic loads. For this purpose a flight test program with 22 flights and specific maneuvers covering low and high frequency control inputs at different airspeeds was performed. In this context, the System-Identification approach is briefly explained and some exemplary results are presented.

FIGURE 19. DLR Discus-2c flight test sailplane [17] FIGURE 21. System-Identification scheme of 7-point loads model 5.2. Results 5.1. System-Identification Approach The following FIGURE 22 and FIGURE 23 show some The conventional System-Identification uses flight data as exemplary System-Identification results for typical input in order to model the aircraft aerodynamic loads. longitudinal and lateral flight test maneuvers. Therefore However, due to the limited information provided by the the time histories show model outputs (red lines) in standard flight test instrumentation only the global comparison to the flight test data (blue lines). The first 8 aerodynamic contributions of wing/fuselage and horizontal plots show control surface deflections and main aircraft tail can be modeled. The objective of this work was to motion parameters like speed, angles of attack and extend the model for providing the capability to simulate sideslip and rotation rates p, q, r. The next 7 plots show local loads at certain positions of wing, horizontal tail and the local aerodynamic lift coefficients for 6 wing segments fuselage. However, this multipoint loads modeling and horizontal tail. requires the availability of local loads measurements as additional observation variables to provide sufficient information for the identification of a significant number of FIGURE 22 depicts a stall approach maneuver combined model parameters. FIGURE 20 shows the wing and with a sequence of elevator doublets. This provides flight horizontal tail cutting loads available at 7 stations after data that cover a large angle of attack, pitch rate and calibrating the overall 46 strain sensors. The notation S, B elevator deflection range. It can be observed that the and T means shear force, bending moment and torque identified 7-point loads model presents a good agreement respectively. By subtracting the structural inertia masses with the measured local aerodynamic loads along the entire angle of attack range. Additionally, these results from the loads measurements local aerodynamic lift (CL), show the efficacy of the implemented stall model used to pitching (Cm) and rolling moment (Cl) coefficients for each account for the wing lift-curve slope reduction at high wing segment and horizontal tail were calculated. CLWR6 for example is the lift coefficient for the right wing segment angles of attack. For analyzing the longitudinal-lateral- from the WR6 load station to the . For the degrees-of-freedom coupling effects and its influence on development of a 7-point flight mechanical model these the local loads, several types of flight maneuvers were measured local aerodynamic coefficients were used like applied to the System-Identification process. One shown in the System-Identification scheme in FIGURE 21 example is the 3-2-1-1 multistep input shown in for finding the best match between measurement and FIGURE 23. It can be observed that the peaks of the local model the DLR software FITLAB [5] was used to estimate spanwise lift coefficient along the wing relate directly to the model parameter values. More detailed descriptions of the aileron excitations. Overall a very good matching the model structure and the identified parameter values between flight test and simulation model is achieved. can be found in [25] and [24].

FIGURE 22. Comparison Measurement – Model: Stall FIGURE 23. Comparison Measurement – Model: 3-2-1-1 approach combined with elevator doublets aileron multistep inputs

Transportflugzeugflügeln. Dissertation, TU-Braunschweig, 6. SUMMARY August 2016. DLR-FB 2016-34. The use case work package of the DLR project iLOADS [10] W. Krüger and T. Klimmek. Definition eines showed the wide range of loads analysis from conceptual übergreifenden Lastenprozesses im DLR-Projekt iLOADS. design to loads flight testing. In Deutscher Luft- und Raumfahrtkongress, 13-15 Sept 2016, Braunschweig (Germany), 2016. For conceptual design a physics base loads and sizing [11] T. L. Lomax. Structural Loads Analysis for process is presented as part of a typical overall aircraft Commercial Transport Aircraft: Theory and Praxis. design process, where normally statistics based or Number ISBN 1-56347-114-0. AIAA Education Series, analytical methods are applied. Furthermore the flexibility 1996. of the structure was not only considered for the structural [12] B. Nagel, T. Zill, E. Moerland, and D. Böhnke. Virtual sizing, but also for the aerodynamic performance analysis. Aircraft Multidisciplinary Analysis and Design Processes – Lessons Learned from the Collaborative Design Project Regarding the loads analysis for a complete aircraft, a VAMP. In in Proceedings of 4th CEAS Air & Space parametric loads and design process is outlined and Conference, 16-19 Sep., Linköping, Sweden, 2013. applied to the DLR HALO. The set-up of a proper structural model with the MONA process is shown as well [13] M. Niu. Airframe Stress Analysis and Sizing. Number as the analysis of selected load cases to be compared to ISBN-13: 978-9627128120. Hong Kong Conmilit Press results from GAC. In order to measure the interface loads LTD, Hong Kong, 1999. between the store and the wing of the HALO with [14] N.N. CS25 - Certification Specifications and mounted PMS carrier, the structural model is also used to Acceptable Means of Compliance for Large Aeroplanes. simulate gust analysis. The results serve as input for pre- http://easa.europa.eu/document-library/certification- tests on the vibration table MAVIS. Eventually first results specifications/cs-25-amendment-14, last checked 2014- of the measured flight loads are shown. 05-20 2013. Amendment 14. [15] D.P. Raymer. Aircraft Design: A Conceptual Finally first results for a loads model for the DLR Discus- Approach. AIAA Education Series, Washington, DC, 2c based on flight test is shown. Two test campaigns of 1989. ISBN 0-930403-51-7. the considerably equipped sailplane with various sensors [16] J. R.Wright and J. E. Cooper. Introduction to Aircraft are a basis not only for a loads model, but for an overall Aeroelasticity and Loads. Number ISBN 978-0470-85840- system identification of the aircraft. 0. Wiley, 2007. [17] Schempp-Hirth Flugzeugbau GmbH, Kirchheim unter SCHRIFTTUM Teck, Germany. Flughandbuch für das Segelflugzeug Discus-2c, 2005. (in German). [1] M. D. Ardema, M. C. Chambers, A. P. Patron, A. S. [18] D. Seider, P. M. Fischer, M. Litz, A. Schreiber, and Hahn, H. Miura, and Mark D. Moore. Analytical fuselage A. Gerndt. Open source software framework for and wing weight estimation of transport aircraft. Technical applications in aeronautics and space. In in Proceedings Report Technical Memorandum 110392, NASA, May of IEEE Aerospace Conference, 3-10 March 2012, Big 1996. Sky, Montana, USA, 2012. [2] P. D. Ciampa, T. Zill, and B. Nagel. A hierarchical [19] J. Sinske, Y Govers, V. Handojo, and W.-K. Krüger. aeroelastic engine for the preliminary design and HALO Flugtest mit instrumentierten Außenlasten für optimization of the flexible aircraft. In In Proceedings of Aeroelastik- und Lastmessungen im DLR-Projekt iLOADS. 54th AIAA/ASME/ASCE/AHS/ASC Structures, Structural In Deutscher Luft- und Raumfahrtkongress 2016, 13-16 Dynamics, and Materials Conference, April 8-11, 2013, Sept. 2016, Braunschweig. DGLR, 2016. (to be Boston, Massachusetts, number AIAA 2013-1820, 2013. published). [3] F. Van Dalen. MDO load analysis and preliminary [20] T.H. Skopinski, Aiken Jr., W.S., and W.B. Huston. sizing. Technical report, Delft University of Technology, Calibration of strain-gage installations in aircraft structures December 1996. for the measurement of flight loads. Technical Report [4] M. J. de C. Henshaw, K.J. Badcock, G.A., Vio, C.B. NACA Report No. 1178, NACA, 1954. Allen, J. Chamberlain, I. Kaynes, G. Dimitriadis, J.E. [21] E. Torenbeek. Synthesis of Subsonic Airplane Cooper, M.A. Woodgate, A.M. Rampurawala, D. Jones, Design. Delft University Press, Delft, 1988. ISBN 90-247- C. Fenwick, A.L. Gaitonde, N.V. Taylor, D.S. Amor, T.A. 2724-3. Eccles, and C.J. Denley. Non-linear aeroelastic prediction [22] F. van Dalen and A. Rothwell. Adas structures for aircraft applications. Progress in Aerospace Sciences, module: User’s manual. Memorandum M-709, TU-Delft, 43:65–137, May-Aug 2007. 1995. [5] DLR - German Aerospace Center. CPACS - Common [23] M. V. Preisighe Viana. Sensor calibration for Parametric Aircraft Configuration Schema, Sept. 2016. calculation of loads on a flexible aircraft. In 16th www.cpacs.de (last checked 2016-09-12). International Forum on Aeroelasticity and Structural [6] DLR - German Aerospace Center. RCE - Remote Dynamics - IFASD 2015, 28 June - 2 July 2015, Saint Component Environment, Sept. 2016. Petersburg, Russia, 2015. www.rcenvironment.de (last checked 2016-09-12). [24] M. V. Preisighe Viana. Multipoint Loads Model for [7] F. M. Hoblit. Gust Loads on Aircraft: Concepts and Flexible Aircraft Health Monitoring in Real Time. Applications, volume ISBN 0-930402-45-2 of AIAA Dissertation, Institut für Flugsystemtechnik, 2016. (to be Education Series. AIAA, 1988. published). [8] R. V. Jategaonkar. Flight Vehicle System Identifica- [25] M. V. Preisighe Viana. Time-domain system identi- tion: A Time Domain Methodology. Progress in fication of rigid-body multipoint loads model. In American Astronautics and Aeronautics, 216, 2006. Reston, VA, Institute of Aeronautics and Astronautics (AIAA), USA. Atmospheric Flight Mechanics Conference, June 2016, [9] T. Klimmek. Statische aeroelastische Anforderungen Washington D.C., USA, 2016. beim multidisziplinären Strukturentwurf von